Publius: The Journal of Federalism Advance Access published June 8, 2009
Government Quality and Vertical
Power-Sharing in Fractionalized States
Nicholas Charron*
Recent empirical studies have demonstrated that ethno-linguistic diversity has a negative relationship with quality of government (QoG). In response to this challenge, states have two broad
options with respect to vertical power-sharing. One, they can attempt to integrate various factions
by adopting a unitary, centralized constitution. Two, they can accommodate minority groups
with ethno-federalism, giving them a degree of regional autonomy. Using numerous QoG
indicators in a cross-sectional regression from 1995 to 2004, the data show that ethno-federalism
outperforms its integrationist rival for each QoG measure employed in the analysis. While
numerous other empirical studies have examined the impact of ethno-federal institutions on
either civil conflict or separatism, this is the first to assess the effects of this institution
on government quality relative to integrationist states.
Much research in the past few decades has been devoted to how states can most
effectively manage ethnic and linguistic diversity within their borders. Broadly
speaking, leaders in democratic countries have two competing strategies that they
can employ to best govern their citizens—integration or accommodation.1 The
former seeks to create a polity, which has one overarching language, culture, and
identity, dismissing the idea that diverse minority ethnic and linguistic groups
should have some degree of privileged political status that distinguishes them from
the majority. The latter tries to accommodate diversity by establishing institutions,
which allow minority groups to participate in politics collectively and coexist with
the majority group. Although there are many types of institutions that represent
the accommodation or consensual school of thought (such as proportional
representation, minority veto, etc.), one such institution is multinational, incongruent, or ethno-federalism, which is a form of federalism that draws the
boundaries of one or more of its provinces around a minority group, giving a
minority group a majority status in at least one province (e.g. Quebec in Canada).
Though scrutiny of this type of accommodation has been intense in the scholarly
*The Quality of Government Institute, Department of Political Science, University of Gothenburg, Sweden;
[email protected]
Publius:The Journal of Federalism, pp.1^21
doi:10.1093/publius/pjp015
ß The Author 2009. Published by Oxford University Press on behalf of CSF Associates: Publius, Inc.
All rights reserved. For permissions, please email: [email protected].
2
N. Charron
literature—especially with respect to civil conflict and separatism as outcome
variables—with states such as Iraq, Sri Lanka, Sudan, and Cyprus currently
considering this institution, more theoretical and empirical work is certainly
needed to understand the impact of this institution on other policy outcome areas.
Although there has been substantial empirical focus on the federalism–
governance nexus, the evidence is mixed on whether decentralization of power
inherent in federalism leads to better governance. While one group of scholars find
that more decentralized states have higher levels of corruption (Treisman 2000;
Gerring and Thacker 2004), others find empirical support for better governance
(Fisman and Gatti 2002; Adesera et al. 2003), and some find no results (Norris
2008). Clearly more work is needed in this area to resolve these findings, but
what previous studies have overlooked is the domestic context in which a federal
constitution exists—in particular whether a relatively heterogeneous country
attempts to accommodate or integrate its diverse groups politically. Further, this
study goes beyond the federal/unitary comparison for all states to analyze the
impact of vertical power-sharing2 political institutions in divided societies—meaning
whether a heterogeneous country has adopted an ethno-federal constitution or not.
This study expands previous research on the effects of ethno-federalism. Large-N
empirical studies on the effects of ethno-federalism have primarily been limited to
analyzing its impact on reducing the likelihood of civil conflict, with mixed results
(Horowitz 1991; Jalali and Lipset 1992: 601–605; Ornstein and Coursen 1992;
McGarry and O’Leary 1993; Gurr 1993 and 2000; O’Leary and McGarry 1995;
Kaufman 1996; Laponce 2003; Brancati 2006). Other studies, which have employed
case study research on states that have recently adopted some aspects of ethnofederalism, have also demonstrated mixed findings. For example, developing
countries such as South Africa (Murray and Simeon 2007) and Indonesia (Bertrand
2007) have moved to more accommodating models and they have created relatively
successful political harmony among diverse groups. Others, such as Singapore and
Malaysia in 1965, Czechoslovakia in 1993, and Yugoslavia in the 1990s, ended in
separatism and/or internal conflict. While state-breakdown or separatism are
certainly salient outcome variables to researchers and policy-makers, as shown in
Hale’s (2004) study on core ethnic group regions, a lack of civil war or separatism
is hardly grounds for claims of success for either ethno-federalism or its rival
(the unitary model) for diverse populations. Thus, I argue, it is incumbent upon
researchers to evaluate the impact of the two rival models (integrationist or
accommodation) using a different standard than state-breakdown. One interesting
empirical question is to what extent political elites in heterogeneous countries manage
the impartial rule of law, do not engage in corruption, and provide quality
bureaucratic services to citizens3 based on whether they have adopted an ethno-federal
system or not. This analysis seeks to evaluate the impact of ethno-federalism in diverse
or fractionalized countries using quality of government (QoG) indicators as outcome
Government Quality and Vertical Power-Sharing in Fractionalized States
3
variables. The aim is to determine whether the vertical power-sharing method of
accommodation has been effective—not only for avoiding civil wars or keeping
diverse groups together within a single state—but also establish whether such a system
produces better governance among diverse groups than integrationist constitutions.
Thus the purpose of this analysis is not to explain why a country has chosen the
accommodation model of ethno-federalism or the cases in which it has broken
down. The primary point is this: given that some states have adopted the ethnofederal model, how do such countries compare—with respect to QoG—to other
fractionalized countries, which have adopted more integrationist-type institutions?
Put another way, does the ethno-federal model—as scholars such as Lijphart
predict—provide elites from various ethnic and linguistic groups the incentives to
govern better, or is QoG higher in states that have sought a more integrationist
model of vertical power-sharing? The literature on power-sharing institutions in
divided societies implies two rival hypotheses regarding this puzzle. Surprisingly,
this question remains untested in the empirical literature.
I test this question on a wide range of states using multiple indicators of QoG in
a cross-sectional sample that ranges from 1995 to 2004. The empirical results
demonstrate two interesting findings. First, the literature has shown that more
heterogeneous (or fractionalized) states, on average, have lower QoG (La Porta
et al. 1999; Alesina et al. 2003), slower economic growth rates (Easterly and Levine
1997; Collier 2000), or greater political instability (Mauro 1995; Annett 1999) than
states that are more homogenous. The data in this analysis support these general
claims and empirical findings. I find that in countries that have higher degrees of
ethno-linguistic fractionalization, the ethno-federal sample shows significantly
higher QoG scores than diverse states that have adopted a unitary/integrationist
model of power-sharing. The results support the notion that the impact of federal
institutions on the QoG is contextualized and nuanced and that, given a diverse
population, certain institutions such as ethno-federalism can encourage better
governance among various groups than its integrationist rival.
The Debate: Integration versus Accommodation
One of the most difficult questions facing societies today is how to respond to
ethnic, linguistic, and religious diversity in ways that promote stability, growth,
democracy, and human rights (Choudry 2008). Multiple ethno-linguistic divisions
might create a greater likelihood for conflict and rent-seeking behavior, which
makes governing more difficult. Broadly speaking, two rival hypotheses are
expected in this well-established literature on power-sharing in divided societies, yet
they have remained untested in a broad, cross-sectional sample of countries.
According to the first theory, societies can choose a Leviathan-type, unitary system
with one set of clear rules for all citizens (Ophuls and Boyan 1992). According to
4
N. Charron
the second theory, they can select a federal constitution where power is divided among
many groups and allow such smaller groups to govern themselves on a wide range of
issues. Recently, a number of scholars have devoted much attention to the effects of
political institutions on QoG (Weingast 1997; Ades and De Tella 1999; La Porta et al.
1999; Treisman 2000; Adesera et al. 2003; Persson et al. 2003; Gerring and Thacker
2004). Although few have focused theoretical arguments on the causal relationship
between ethno-federal or integrationist institutions on QoG in divided societies,
there is a wealth of literature implying the effect, which I draw on below.
The Arguments for Accommodation
Since the writings of James Madison, proponents have argued that institutions such
as regional representation serve as a political outlet for multiple distinct ethnic,
religious, or linguistic groups in a society. In the modern-day literature, many
proponents of consociationalism have argued that identities in a society are often
too resilient to overcome with integrationist institutions—thus, a state with sizeable
minority groups should, in fact, accommodate these factions through institutions
such as proportional representation, minority veto, and—of particular importance
to this study—ethno-federalism (see McRae 1974; Lijphart 1977, 1985, 1999;
Kymlicka 1995; Watts 1999a, b).
Elite Bargainingç‘Trickle Down Tolerance’
While large-N comparative studies of ethno-federalism outside of the civil conflict
literature are rare, some scholars have certainly theorized about the effect of ethnofederalism and cultural autonomy on QoG in fractionalized states. Lijphart (1984,
1985, 1997) postulates that distinct ethnic and linguistic groups within a heterogeneous society should be allowed to have corresponding political sub-national
governments in which to conduct a number of policy avenues autonomously.
In heterogeneous societies, the key to tolerance and cooperation—which lead to
impartial laws, legitimacy, and a better-functioning government—are elites, who
can either send hostile or cooperative signals to their constituencies regarding
relations with other groups. Lijphart suggests that the best way in which to obtain
elite consent necessary for cooperation is not by forcing them—particularly for
minority elites—to fight for power in a unitary state, but by offering them an
automatic share of power concentrated in a sub-national region. He thus offers
a top-down/elite-driven theory in which leaders of diverse social groups that have
their own legitimate sub-national governments work together, and with the
majority group at the national level. Provided that local elections and a certain
degree of provincial fiscal legislation are allowed, the cultural and linguistic
identities of minority groups remain intact and protected by the government.
For example, the important policy areas include education curriculum and right to
Government Quality and Vertical Power-Sharing in Fractionalized States
5
teach a different language and history in schools, taxation levels, and local security
and justice laws. In order to preserve their status as regional leaders with a degree
of legitimate power to lose, such elites persuade their voters to accept the status
quo of the federal system. With the consent of the minority and majority-group
elite members, the legitimacy and content of the systems among the citizenry
begins. Over time, a stable, well-governed country (relative to unitary states) is
expected. For example, in a well-argued policy prescription designed to influence
the constitution of Iraq, McGarry and O’Leary (2007) maintain that the consensual,
ethno-federal model would be the most appropriate—not only to avoid future civil
war, but to give each group a stake in politics and to ‘‘create incentives for an
executive that is responsible to different regional constituencies’’ (McGarry and
O’Leary 2007: 693). In sum, elites of multiple ethnic or linguistic groups will have
much less incentives to fight the central government and will instead be more
persuaded to coalesce together nationally, while simultaneously providing better
governance locally to ethnically or linguistically similar people, as opposed to being
forced into an integrationist arrangement—where they are at best under-recognized
by a larger, majority group and, at worst, targeted and discriminated by laws
and bureaucratic services that favor the majority group in power.
Social Capital, Trust, and Legitimacy
Heterogeneous societies have more actors vying for resources and thus increase the
likelihood for collective action problems and have problems with respect to social
trust in government relative to more homogenous states. Some scholars have noted
that one possible way a state can aid in overcoming trust-deficit and collective
action dilemmas is through the promotion of social capital (Ostrom 1990; Putnam
1994; Mazzone 2001) and general trust in society by providing minorities a stake in
the democratic society (Posen 1993; Fearon 1996), thus leading to impartial
institutions and better overall satisfaction in governance (Rothstein 2007).
A constitution that allows for sub-national government is one where multiple
interests and ethnic groups can participate in politics while not vying for power in
one capital city. Majority groups send a credible signal to minorities that their
political preferences will be heard both in the region in which they live and at the
central level, most often through an upper chamber. Conversely, unitary states with
diverse populations allow for the potential domination of one or narrow-interest
groups, which might leave out minority constituencies from policy-making.
Federalism provides such a forum for citizens to ‘‘actively increase opportunities
for engagement in government by additional groups of citizens, thereby enhancing
social capital’’ (Mazzone 2001: 42). Gurr (1993) presents empirical evidence
showing that divided states that allow for cultural autonomy reduce the likelihood
of protests and rebellions among minority groups. Citizens can more actively
6
N. Charron
participate in social and political networks knowing that they can participate at
a more local level with others more like themselves, fostering more trust and
encouraging collective action.
Encouraging Intra-Ethnic Responsibilities while Increasing Minority Protection
Another way that the accommodationist model can help to alleviate potential
collective action problems among diverse groups in a single country is simply
through dividing up the responsibility of governance (Kymlicka 1995, 1998;
Watts 1999a). As Olson (1965) points out, smaller group size is easier to manage in
order to produce the common good—in this case, quality governance for all
diverse groups. By reducing the number of participants from governing in one
central level, specific policy preferences and goals can be more easily identified and
smaller groups—in this case ethnic or linguistic minorities—have a greater
incentive to take responsibility for themselves rather than buck-passing—relying
on the majority group in power at the center to govern on their behalf.
This type of arrangement essentially builds privileged local elite classes and
increases democratic participation in minority regions, whereby leaders of
such groups have stronger incentives to take on more responsibility of providing
better governance and economic goods to their regional constituents (see Olson
1965: 48–50).
By providing cultural and linguistic autonomy to various minority groups
within a diverse society, provincial governments constrain potential abuses of
power at the central level directed at minority groups by offering multiple groups
in society the authority to maintain their linguistic or ethnic integrity, as in the
case with the Francophone population in Canada residing in Quebec (Watts
1999a). Federal arrangements thus reduce the likelihood of potential abuses of
power by the majority group at the central level by shielding the minority groups
from domination.
Such a constitutional design gives individuals and minority groups protection
from the center while simultaneously interlocking decision-making on salient policy
areas such as education, where the provincial governments can influence the
decisions made at the central level. Such an arrangement can lead to greater
satisfaction and less incentive by central authorities to seek rents at the expense of
other groups because there are fewer opportunities to do so, relative to a more
centralized, unitary state (Goodin 1996). Simultaneously, it puts accountability on
leaders in minority and majority regions alike, to produce quality outputs in
their policy spheres, knowing that they cannot blame certain policy failures on
the other group during their provincial elections.
The debate over whether to have an integrationist or accommodation institution
implies that the impact of fractionalization on QoG will be conditioned by either
Government Quality and Vertical Power-Sharing in Fractionalized States
7
ethno-federalism or the more integrationist strategy. The following hypothesis will
thus be tested in this analysis:
H1: With respect to vertical power-sharing, ethno-federalism is more likely
to produce higher levels of QoG in divided societies than more integrationist
institutions.
The Arguments for Integration
Problems with ‘‘Asymmetrical Laws’’
Theoretical arguments for a more integrationist institutional model agree that
building one nation, founded on a common culture and language is of primary
importance in constitutional design (Brass 1991; Horowitz 1991, 2000; Barry 2000;
Reilly 2001). Integrationist institutions are designed with the intent to build
impartial and equal citizenship among a polity, which would lead to relatively
better rule of law than by accommodating minorities. Such scholars argue that
accommodationist practices can institutionalize and exacerbate differences among
various diverse groups that they are designed to manage (Horowitz 2000). These
arrangements can be particularly harmful when minority regions obtain greater
representation than their population would warrant under a unitary system (Yash
Ghai 2002). This is because, if the balance of power is perceived as being tilted
unfairly, it could lead to resentment by the majority and encourage poorer
bureaucratic services to minority groups and bring about confusion regarding
decision rules, which also potentially exacerbates corruption and poor government
quality (Gerring and Thacker 2004). Thus allowing special privileges to certain
groups in the form of autonomy or federalism, which draws territorial boundaries
around certain minority groups, undermines the goal of ‘‘one nation’’. Additional
special privileges and/or asymmetric representation for particular ethnic or
linguistic groups can lead to a ‘‘lack of cohesion and problems of governability’’
(Yash Ghai 2002: 159). Integrationist institutional designs with respect to vertical
power-sharing would thus be either unitary or a form of federalism that is not
designed around any ethnic or linguistic criteria (i.e. the United States or
Germany).
The ‘‘Incompetent Lower-Level Bureaucrat’’ Argument
Some emphasize that federal systems in general are less likely to attract competent
policy makers and bureaucrats at the regional and local levels because the benefits
for such positions would be considerably less rewarding than similar positions at
the central level (Tanzi 1996, 2001). Specifically, some scholars take issue with the
notion that decentralized/federal institutions lead to better political outcomes in
8
N. Charron
fractionalized societies. Such scholars argue that corruption may be greater at the
provincial and local levels, possibly due to easier access to such policy makers when
sought out by interest groups. This scenario creates an environment where
corruption and bureaucratic inefficiency are more likely at the regional levels, thus
leading to poorer quality government in decentralized countries (Ghai and Regan
1992; Prud’homme 1995; Tanzi 1996; Brueckner 1999). Further, in developing
states, which may have poorly equipped regions led by minority ethno-linguistic
groups, the fiscal and political responsibilities can be overwhelming, leading to
massive inefficiencies in the bureaucracy (Yash Ghai 2002).
Confusion regarding Accountability
Another argument along these lines is whether a federal structure could lead to less
overall accountability if political decision-making is made at multiple levels of
government. On one hand, voters have a more difficult time assessing blame for
poor-quality governance, thus providing some with more of an incentive to extract
rents. On the other hand, bureaucrats and policy makers could theoretically be less
coordinated among the levels of government and, with each level being equally
corrupt; there could be a disproportionate amount of rent-seeking, relative to more
centralized states (Shleifer and Vishny 1993). Such confusion among levels can lead
to finger-pointing and buck-passing from one region/ethnic group to another,
which gives rent-seeking opportunities for politicians and bureaucrats. This can be
especially prevalent in ethno-federal states where relations between two or more of
the diverse groups are antagonistic, encouraging the blame of the ‘‘other’’ for poor
governance, while simultaneously engaging in corruption.
Additionally, ethno-federalism, like mono-federalism, would add an additional
‘‘veto point’’ (Tsbelis 2000). Because the status quo is less likely to change as
a function of the number of veto points, this could on the one hand be helpful to
ensure impartial and beneficial laws and policies to most citizens. However, as
Gerring et al. (2007: 8–10) argue, federalism can also make changing a corrupt,
sub-optimal status quo more difficult.
Questionable Motivation of Elites
Finally, some critics of ethno-federalism question the consociational assertion that
given multi-ethno-linguistic minority rule in sub-national regions and guaranteed
representation at the central level, elites from such diverse groups will be more
likely to tolerate and negotiate with one another to produce a moderate political
environment (Reilly and Reynolds 1999; Horowitz 2002). As Horowitz (1997, 2002)
points out, the motivations of elites toward other ethnic/linguistic groups in an
ethno-federation, with respect to tolerance and compromise, may or may not be
less ethnocentric than the masses they represent. In some cases, elites that do in fact
Government Quality and Vertical Power-Sharing in Fractionalized States
9
choose to compromise might pay a political price from their own base, being
branded a ‘‘sellout’’ (Horowitz 2002: 21). In cases where the incentives for elites to
cooperate among diverse groups are low, more integrationist political institutions
that encourage broad coalitions among diverse constituencies—rather than
allowing and encouraging them to reinforce ethno-linguistic differences—are
argued to be more beneficial for cooling potential tensions and conflicts, which
certainly threaten good governance and the impartiality of the rule of law.
The preceding literature leads to the following hypothesis:
H2: With respect to vertical power-sharing, integrationist institutions are more
likely to produce higher levels of QoG in divided societies than ethno-federalism.
Research Design
The research design consists of a cross-sectional, multivariate regression for each of
the four dependent variables. This allows for a number of control variables to be
included to account for rival hypotheses of QoG’s determinants. Because the
primary independent variables are largely time-invariant, the cross-sectional
approach is appropriate for this analysis,4 including the interaction term. The
sample is stratified, and only states that are at least partially democratic according
to Polity IV are included.5 This is relevant because both the accommodationist and
integrationist proponents discuss the bargaining relationship among elites and
citizens of diverse groups in a democratic setting. Thus dictatorships that are coded
as ‘‘ethno-federal’’ (such as Nigeria or Ethiopia in this case) are dropped from the
analysis, along with all non-democratic ‘‘integrationist’’ countries. Concerning the
timeframe, I use country nine-year averages for all available data in the analysis.
Following the advice of Ray (2005), I provide a baseline for the model, using only
the key independent variables individually, before reporting the full models, which
include control variables and the interaction term. Finally, I take a stratified
sample that includes only developing, or newly federal states in the analysis.6
Presentation of Data and Methods
Dependent Variables
To test the full scope of the effect of power-sharing institutions on government
quality, I employ four dependent variables in the study. According to the Quality of
Government Institute,7 the three core areas that serve as empirical proxies for
government quality as a concept are low levels of corruption, high levels of
bureaucratic quality and effectiveness, and strong democratic institutions and
participation among a country’s citizens. Due to admittedly problematic data at
times with such concepts as corruption and level of bureaucratic quality, I utilize
10
N. Charron
multiple proxy measures to account for any potential aberrations in the empirical
findings of any single measure.
Two of the dependent variables employed here are taken from the Political Risk
Services Group’s (PRS) International Country Risk Guide (ICRG). The PRS Group,
a think tank specialized in economic and political risk assessment internationally,
has published monthly data for business and investors on over 140 countries since
1980. The primary variable, representing all three components of QoG is the mean
value of the ICRG variables ‘‘Corruption,’’ ‘‘Law and Order,’’ and ‘‘Bureaucracy
Quality,’’ scaled 0–1. Additionally, I take the measure of ‘‘corruption’’ separately,
which accounts for ‘‘excessive patronage, nepotism, job reservations, ‘favor-forfavors,’ secret party funding, and suspiciously close ties between politics and
business.’’8 The data in the analysis has a finite range from ‘‘0’’ to ‘‘1’’, with higher
scores indicating lower levels of perceived corruption. It has also been used by
other recent empirical studies (Ades and Di Tella 1999; Fisman and Gatti 2002;
Persson, Tabellini, and Trebbi 2003,).
The third measure employed here comes from Transparency International;
a non-partisan organization that has created the ‘‘Corruption Perceptions Index’’
(CPI), intended to capture corruption. The CPI score measures the ‘‘perceptions of
the degree of corruption as seen by business people, risk analysts, and the general
public’’.9 The fourth measure is intended to capture ‘‘government effectiveness’’ (or
‘‘bureaucratic quality’’) from the World Bank Governance Indicator (Kaufmann,
Kraay, and Mastruzzi 2007). Specifically, they code this variable on the basis of ‘‘the
quality of public service provision, the quality of the bureaucracy, the competence
of civil servants, and the independence of the civil servants from political pressures
and the credibility of the government’s commitment to policies’’.10 A more detailed
description of these variables can be found in the Supplementary Appendix A1.
Independent Variables
Ethno-Federalism
For the purposes of this analysis, ethno-federalism is defined as possessing three
features: (1) sub-units that are defined territorially and identified with an ethnic
and/or linguistic minority; (2) dual sovereignty, where policy-making is divided
between a center and its sub-units and; (3) a center-sub-unit relationship that is
marked by coordination and autonomy (see Riker 1964; Stephan 1999; Filippov,
Ordeshook, and Shvestsova 2004; Bunce 2004).
I take a contemporary list of federal systems from the Forum of Federations11
and cross-check these data with the Database of Political Institutions (DPI, Kiefer
et al. 2005). The countries that are selected as ‘‘ethno federations’’ are checked with
previously published studies on this topic (Laponce 2003; Verney 2004; József
2005). The measures important to this study are of course whether a state has
Government Quality and Vertical Power-Sharing in Fractionalized States
11
Table 1 List of federal states
Argentina
Australia
Austria
Belgium
Bosnia & Herzegovina
Brazil
Canada
Comoros
Ethiopia
Germany
India
Malaysia
Mexico
Micronesia
Nigeria
Pakistan
Russia
St. Kitts & Nevis
South Africa
Spain
Switzerland
United Arab Emirates
United States
Venezuela
Note: States considered to be ‘‘ethno-federations’’ are listed in italics.
Source: Forum of Federations, DPI (2004), Watts (1999b).
provincial legislatures and if, in fact, they are locally elected rather than appointed
by the central government, and whether or not such provincial governments have
‘‘extensive taxing, spending, or regulation authority.’’12 These elements follow the
standard of what is considered a decentralized or federal country.13 I thus generate
a dichotomous indicator for whether a country has an ethno-federal (‘‘1’’) or an
integrationist political structure (‘‘0’’), excluding those considered ‘‘monofederal’’
(Juhasz 2005) due to the specific prediction that ethno-federal constitutions can aid
in better governance in ethnically diverse states.14 A full list of the federal states
is listed below in table 1, with ethno-federations in parentheses.
Fractionalization
The second important aspect of this study pertains to whether or not
accommodationist/integrationist institutions have a lesser or greater effect
depending on if a country is heterogeneous or homogenous with respect to
ethnicity and/or language. I employ a common measure to capture this effect—
ethno-linguistic fractionalization—from the study of Alesina et al. (2003). Values
are calculated as one minus the Herfindahl index of ethnic,15 linguistic, and
religious group shares of the total annual population. The figure essentially
represents the probability that two randomly selected individuals from a country
will belong to two different ethnics, linguistic, or religious groups. The index ranges
from ‘‘0’’ to ‘‘1,’’ with higher values equating to higher fractionalization. Though
I anticipate that higher fractionalization will relate negatively with quality of
government, the key to this analysis is that I intend to demonstrate the contextual
impact of accommodationist/integrationist institutions, meaning that when
heterogeneity is high, I test which institutions are associated with better
government quality. To maximize the full amount of variance in the sample, I
construct an interaction term with the ethno-federal variable and ethno-linguistic
fractionalization to test for this effect (ethno-fed.*fractionalization).
12
N. Charron
Control Variables
Though I try to be as parsimonious as possible, some control variables are needed
in the model to account for rival hypotheses. First I include the size of a country by
accounting for its population. Such a measure is useful in avoiding spurious
correlation between larger countries with more diverse populations adopting federal
structures to best provide for their citizens. I also include GDP per capita from the
Penn World Tables (Heston et al. 2002) to account for economic development.
I take the natural log of country’s GDP. Furthermore, a number of empirical
studies have found lower degrees of corruption in states with a free press (Ahrend
2002; Brunetti and Weder 2003; Lindstadt and Naurin 2005). I use the Freedom
House measure of press freedom, which scores states on a scale of 0–100. I reverse
the index so that higher scores indicate higher levels of press freedom. Finally,
because some empirical studies have found systematic differences for QoG variables
when testing for certain regions, such as the Middle East (Norris 2008), Latin
America (Rodrik 2000; Treisman 2000), and sub-Saharan Africa (Treisman 2000),
a number of area dummies are included in the analyses. All are dichotomous (0/1)
indicators as to whether or not a country is located in Africa, Latin America, or the
Middle East. Again, a more detailed description of all variables used in the analyses
is located in the Supplementary Appendix A1.
Results
Table 2 shows the six different empirical models all designed to test the two rival
hypotheses with countries that rank above a five on the Polity IV.16 Model 1 is a
test of the basic relationship of the two primary variables on the PRS Group
measure of QoG, which includes rule of law, corruption, and bureaucratic quality.
Upon first glance, the coefficient for ethno-federalism indicates that the
accommodation model has a positive, but insignificant effect on QoG, while the
impact of fractionalization is clearly negative, a result found commonly in
the literature. Model 2 employs the same dependent variable, yet adds more nuance
to the analysis with the interaction term and the controls for population, free press,
economic development as well as area dummies. While in model 1, the impact of
ethno-federalism was insignificant, here in model 2 we observe that its impact on
QoG is highly contextualized—the interaction term is positive and significant at the
95 percent level of confidence and in cases of high fractionalization, the effect of
the ethno-federal model is positive and significant as compared with the
integrationist constitutions (see figure 1 and table 3). However, looking at the two
individual constituent variables individually, the coefficients might be misleading at
first glance with the interaction term present. The result for fractionalization
represents the effect of this variable only for integrationist/unitary states, which is
negative and insignificant. The coefficient for ethno-federalism, which is negative
0.124 (0.121)
0.410 (0.087)***
Primary independent variables
Ethno-Federalism
Fractionalization
Fed*Fractionalization
79
0.00
0.22
Obs.
Prob.4F
R2
79
0.00
0.83
0.005 (0.006)
0.061 (0.047)
0.093 (0.011)***
.172 (0.098)*
0.293 (0.108)**
0.068 (0.048)
0.561 (0.200)***
Full modela
2
85
0.00
0.80
0.003 (0.07)
0.77 (0.53)
1.24 (0.13)***
5.82 (1.03)***
2.58 (0.75)***
0.35 (0.60)
5.02 (1.55)***
Full modela
3
105
0.00
0.81
0.103 (0.021)***
0.531***
0.473 (0.041)***
4.65 (0.310)***
0.802 (0.368)**
0.229 (0.212)
1.63 (0.805)**
Full modela
4
79
0.00
0.64
0.004 (0.008)
0.019
0.094 (0.014)***
0.104 (0.151)
0.360 (0.114)***
0.093 (0.078)
0.681 (0.241)***
Full modela
5
ICRG - C
77
0.00
0.64
0.065 (0.020)***
0.664***
0.307 (0.038)***
3.29 (0.317)***
0.571 (0.179)***
0.111 (0.211)
1.07 (0.354)***
Developing
States onlyb
6
Note: OLS regression with robust standard errors (in parentheses). Limited to states coded as above a ‘‘5’’ according to Polity for a majority of the years
from 1996 to 2004. Developing states are outside the OECD.
a
Indicates that model also contains area dummies for Middle East, Africa, and Latin America.
Dependent variable is ICRG Quality of government (rule of law, control of corruption and bureaucratic quality). Alternative indicators are Transparency
International (corruption), World Bank (‘‘government effectiveness’’) and ICRG (corruption only).
b
‘‘Developing-States’’ model uses World Bank data for maximum number of observations.
*p50.10, **p50.05, ***p50.01.
.798 (20.47)***
Population (log)
Press Freedom
GDP Per Capita (log)
Constant
Control variables
Baseline
1
Trans. Int’l
(ICRG QoG)
World Bank
Alternate measures of QoG
QoG
Table 2 Effect of vertical power-sharing in democratic and partially democratic states on quality of government (QoG)
Government Quality and Vertical Power-Sharing in Fractionalized States
13
N. Charron
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Marginal Effect of Ethnolinguistic Fractionalization
14
0
1
0 =Unitary 1=Ethno-Federal
Marginal Effect of Ethnolinguistic Fractionalization
95% Confidence Interval
Figure 1 Marginal effect of ethno-linguistic fractionalization on QoG, dependent variable: ICRG
QoG scores.
Table 3 Predicted level of QoG in ethno-federal states at different levels of ethno-linguistic
fractionalization
Level of fractionalization
Predicted effect
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.195
0.151
0.106
0.061
0.017
0.027
0.072
0.117
0.161
0.206
(0.074)**
(0.061)**
(0.048)**
(0.038)
(0.031)
(0.031)
(0.036)**
(0.047)**
(0.059)***
(0.072)***
95% Confidence Interval
0.343 to 0.048
0.272 to 0.029
0.202 to 0.010
0.137 to 0.014
0.079 to 0.045
0.033 to 0.089
0.001 to 0.147
0.022–0.212
0.042–0.281
0.061–0.352
and significant, is only interpretable when ethno-fractionalization is ‘‘0’’, which
would defeat the purpose of any ethno-federal boundaries within a country and
of course does not occur in the sample. Thus one must take into account
the interaction coefficient to understand the effects of ethno-federal states in
heterogeneous countries relative to unitary states (see Figure 1).
Government Quality and Vertical Power-Sharing in Fractionalized States
15
Moving to alternative indicators of QoG in models 3–5, we can see that the
results are extremely robust throughout the empirical tests. Model 3 employs
Transparency International data on perceived corruption, and again, shows that the
interaction between ethno-federal states and fractionalization is strong, with this
accomodationist model performing significantly better than its integrationist
counterpart for diverse states. Again, the differences between the accommodationist
and integrationist models widen as ethno-linguistic fractionalization increases.
The results are robust irrespective of the indicator. The World Bank data and
the disaggregated PRS Group indicator, measuring corruption perception only,
both show strong interactive effects between the ethno-federal variable and
fractionalization. Finally, model 6 excludes all OECD states in order to test if the
results in models 2–5 were driven by the more ‘developed’ democracies in the
sample. The developing-state model shows that the results are not sensitive to
the OECD states. The interaction term is strongly significant at the ninety-nine
percent level of confidence and is almost twice as large as the ethno-federal dummy
coefficient, indicating that the model predicts a positive and significant effect of
the accommodationist model at higher levels of fractionalization.
Briefly, the effects of the control variables are mostly as anticipated. Economic
development is by far the most robust determinant of QoG in Table 2. Roughly
speaking, a one standard deviation increase in GDP per capita is predicted to
improve QoG measures by slightly less than half of a standard deviation of each of
the four indicators (see Supplementary Appendix A1 for summary statistics). Press
freedom and population both are positive and significant at the 99 percent level of
confidence in models 4 and 6, which use World Bank data, yet fail to reach even
the 90 percent level of significance in models 2, 3, and 5.
Figure 1 shows visually the marginal effect of fractionalization on QoG
conditioned by vertical power-sharing with the standard errors for the effect of the
modifying variable, in this case the presence or absence of ethno-federalism.17 The
graph was run using the estimates from the ICRG data from model 2 in Table 2
with the full sample of democratic and partially democratic states. The solid sloping
line indicates how the marginal effect of ethno-fractionalization changes depending
on whether a state is ethno-federal or not. The two dash-lines are ninety-five
percent confidence intervals around the marginal effect, which allow us to
determine the conditions under which fractionalization has a statistically significant
effect on QoG—which is whenever the upper and lower bounds of the confidence
interval are both above (or below) the zero line. The two markers indicate the
effect of fractionalization when the ethno-federal variable is equal to ‘‘0’’ and ‘‘1’’.
Since this is a dichotomous constituent variable, the line in between the two
markers is for all practical purposes superfluous. The joint-effect of ethno-linguistic
fractionalization in countries, which have adopted a more integrationist design for
vertical power-sharing (mainly unitary systems) has somewhere from a null-impact
16
N. Charron
to a slight negative impact on QoG, yet the ninety-five percent confidence intervals
crosses the ‘‘0’’ line, meaning we cannot substantively interpret this effect.
However, the data show that compared with the unitary model, the joint impact
of fractionalization in an ethno-federal state on QoG is strongly positive.
Furthermore, the confidence interval around the point indicates the marginal effect
is statistically distinguishable from zero.
In order to elucidate at what point ethno-federal constitutions actually have a
positive effect on QoG in fractionalized states, I calculate the predicted effects in
table 3 according to the same model (model 2 in table 2). Table 3 also provides
some context to the results in each of the models 2–6 in table 2, which show
that ethno-federal states actually reduce QoG in cases of low fractionalization.
The predictions of such values can be observed in table 3—in cases where
fractionalization is low, 0.1 to 0.3, the effect of ethno-federalism on QoG is shown
to be negative and significant. However, in the data, the only ethno-federations that
have a fractionalization score lower than 0.5 are Russia and St. Kitts and Nevis.
Yet Russia does reach the threshold of democracy in model 2 and St. Kitts and
Nevis has no recorded score according to the PRS Group. Thus when we consider
only ‘‘real world’’ cases in the actual sample, the predicted effects of eligible ethnofederations fall between about 0.51 and 0.85 with respect to ethno-linguistic
fractionalization. The results from table 3 show that ethno-federalism has no
statistically significant effect in cases where fractionalization is 0.66 or lower. Yet
from 0.67 to 0.69, the effect is positive and significant at the 90 percent level, from
0.70 to 0.78 it is significant at the ninety-five percent level of confidence, and from
0.79 onward the effect is positive and significant at the ninety-nine percent
confidence level. Overall, the more diverse the state, the stronger the predicted
benefit that an ethno-federation produces in terms of QoG.
Conclusion
How to devise political institutions and share power in divided societies ‘‘is one of
the most difficult and important issues of contemporary politics.’’ (Choudry 2007:
1). This analysis has contributed in evaluating two contrasting constitutional
designs for heterogeneous societies—accommodationist and integrationist vertical
power-sharing—focusing on their performance with respect to government quality.
The results demonstrate that for a number of different QoG indicators, ethnofederations outperform integrationist constitutions at the aggregate level in cases
where ethno-linguistic fractionalization is relatively high. Though little empirical
work has been done previously in a large-N design, the theoretical literature
predicts two competing hypotheses on the expected effect of ethno-federalism in
divided societies. With many current transitional states today having high degrees
of ethnic and linguistic diversity, this study serves as a valuable first step in helping
Government Quality and Vertical Power-Sharing in Fractionalized States
17
policymakers and scholars alike aid such transitioning countries with critical
decisions regarding political institutions. While it would be a daunting task for a
government to alter its linguistic or ethnic demographic make-up, institutions on
the other hand can be designed and/or changed to meet the unique needs of
a diverse (or homogenous) citizenry. Thus, one critical way in which researchers
and policy-makers can assist diverse countries with better governance is to
encourage institutions that have worked well for similar countries in the past. With
the caveat in mind that some ethno-federations of course have in fact separated—
Czechoslovakia, Yugoslavia, USSR, for example—the empirical results demonstrate
for those states that have been ethno-federal and at least partially democratic, they
have clearly outperformed the more integrationist states, irrespective of the QoG.
After a certain degree of ethno-linguistic fractionalization (roughly 0.66 predicted
by the model) ethno-federalism showed to significantly outperform more integrationist institutions, such as the unitary model with respect to QoG indicators.
Its effect was predicted to strengthen with higher degrees of fractionalization.
Ethno-federal constitutions have drawn much attention and controversy in
the theoretical, case study, and civil conflict literature; yet previous large-N
examinations with outcome variables such as government quality are underexplored. While studies that evaluate whether or not ethno-federations are more
likely to erupt in civil conflict or succession are important, it is also critical that we
go beyond whether or not ethno-federations are simply stable systems in divided
societies. Whether or not such constitutional designs lead to better governance,
economic performance, or other social outputs is critical in designing constitutions
for ethnically and linguistically diverse countries.
Supplementary Data
Supplementary data can be found at www.publius.oxfordjournals.org.
Notes
I would like to thank Anette Ahrnens, Matt Golder, Paul Hensel, Sören Holmberg, Victor
Lapuente, Naghmeh Nasirtousi, Bo Rothstein, Dale Smith, Jan Teorell, and the anonymous
editors at Publius: The Journal of Federalism for their help and comments on this article.
1. These two concepts are also referred to as ‘‘majoritarianism’’ and ‘‘consentualism’’
(see Lijphart 1984; Choudry 2007; McGarry and O’Leary 2007).
2. Meaning the way in which the central government shares power with regional subunits—i.e. provinces, states, Länder, cantons, etc.
3. These three indicators are the primary indicators of ‘quality of government’ (QoG)
according to the Quality of Government Institute in Gothenburg, Sweden.
18
N. Charron
4. I would like to thank a reviewer for recommending this approach as opposed to using
the data in panel form.
5. I would like to thank a reviewer for this suggestion. Polity IV description is found in the
appendix along with each country’s respective average score for the time period in the study.
6. I would like to thank the reviewers at Publius for this recommendation.
7. The Quality of Government Institute, founded by Bo Rothstein and Sören Holmberg, is
based at the University of Gothenburg in Sweden. It is a research organization
committed to understanding ‘‘how and under what circumstances high quality political
institutions can be created.’’ For more information, see: http://www.qog.pol.gu.se/
8. See http://www.prsgroup.com/ICRG_Methodology.aspx
9. See www.trancparency.org
10. See www.worldbank.org/wbi/governance/pubs/govmatters4sra.html
11. The Forum of Federations is a ‘‘non-profit, international organization based in Ottawa,
Canada’’ and can be found online at: http://www.forumfed.org/en/
12. See http://econ.worldbank.org
13. These indicators are selected according to Riker’s definition, where there are two key
components to a federal state. First, there must exist two or more levels of government
(central and provincial, etc.). Second, both (and/or all) levels must have constitutionally or
legally recognized policy responsibilities that are independent of other levels of the
government.
14. As for ‘monofederations’, such as the United States and Germany, which do not draw
provincial boundaries around ethnic or linguistic lines, I run models with such states
included in the ‘integrationist’ side (non ethno-federal), and robust-check models
(located in the Supplementary Appendix A1) with such states are simply omitted from
the analysis. However, the differences in the results are inconsequential.
P
15. The standard Herfindahl index is: H ¼ ni¼1 si2 which simply calculates a society
composed of ‘i’ 2 different ethnic groups and ‘Si’ indicates the share of group ‘i’ in
the total population.
16. Additionally, I run robustness checks to include polity IV scores of any country higher
than 4. The results weaken slightly with the inclusion of some less democratic/
‘‘transitional’’ ethno-federal states, yet do not change substantively.
17. I am grateful to Brambor, Clark, and Golder (2006) for posting their Multiple Interaction
Models explanation guide in STATA online. The code for running this graph was taken
from the following Web site: http://homepages.nyu.edu/mrg217/interaction.html
References
Adserà, Alı́cia, Carles Boix, and Mark Payne. 2003. Are you being served? Political
accountability and quality of government. The Journal of Law, Economics, & Organization
19: 445–490.
Government Quality and Vertical Power-Sharing in Fractionalized States
19
Aghion, P., A. Alesina, and F. Trebbi. 2002. Endogenous political institutions. Working
Paper, National Bureau of Economic Research.
Alesina, A., R. Baqir, and W. Easterly. 1999. Public goods and ethnic divisions. Quarterly
Journal of Economics 114: 1243–1284.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and
Romain Wacziarg. 2003. Fractionalization. Journal of Economic Growth 8: 155–194.
Annett, A. 1999. Ethnic and religious division, political instability, and government
consumption. Photocopy, International Monetary Fund.
Brambor, Thomas, William Clark, and Matt Golder. 2006. Understanding interaction
models: Improving empirical analyses. Political Analysis 14: 63–82.
Brancati, David. 2006. Decentralization: Fueling the fire or dampening the flames of ethnic
conflict and secessionism? International Organization 60: 651–685.
Brunnetti, Aymo, and Beatrice Weder. 2003. A free press is bad for corruption. Journal of
Public Economics 87: 1801–1824.
Collier, P. 2000. Ethnicity, politics and economic performance. Economics and Politics 12:
225–45.
Easterly, W., and R. Levine. 1997. Africa’s growth tragedy: policies and ethnic divisions.
Quarterly Journal of Economics 111: 1203–1250.
Fearon, J., and D. Laitin. 1996. Explaining inter ethnic cooperation. American Political
Science Review 90: 715–729.
Fisman, Raymond, and Roberta Gatti. 2002. Decentralization and corruption: evidence
across countries. Journal of Public Economics 83: 325–345.
Gerring, John, and Strom C. Thacker. 2004. Political institutions and corruption: The role of
unitarism and parliamentarism. British Journal of Political Science 34: 295–330.
Ghai, Yash. 2002. Constitutional asymmetries: communal representation, federalism, and
cultural autonomy. In The Architecture of democracy: constitutional design, conflict
management, and democracy, ed. Andrew Reynolds, 141–170 (Oxford: Oxford University
Press).
Ghai, Y. P., and A. J. Regan, 1992. The law, politics and administration of decentralisation in
Papua New Guinea. Boroko: National Research Institute.
Gurr, Ted Robert. 1993. Why minorities rebel: a global analysis of communal mobilization
and conflict since 1945. International Political Science Review 14: 161–201.
———. 2000. Peoples versus states: Minorities at risk in the new century. Washington, DC:
United States Institute of Peace Press.
Horowitz, D.L. 1991. A democratic South Africa? Constitutional engineering in a divided
society. Berkeley: University of California Press.
Horowitz, Donald. 1997. Encouraging electoral accommodation in divided societies,
in Electoral systems in divided societies: The Fiji constitution review, ed. Brij Lal and
Peter Larmour, 21.
N. Charron
20
———. 2000. Constitutional design: An oxymoron? NOMOS Designing Democratic
Institutions 42: 253–284.
———. 2002. Constitutional design: Proposals versus processes. In The Architecture of
Democracy: Constitutional Design, Conflict Management, and Democracy, ed. Andrew
Reynolds, 15–36 (New York: Oxford University Press).
Jalali, R., and S.M. Lipset. 1992. Racial and ethnic conflicts: A global perspective. Political
Studies Quarterly 107: 585–60.
József, Juhász. 2005. Ethno-federalism: Challenges and opportunities. Med-unarodni problemi
57: 245–263.
Kaufman, Chaim. 1996. Possible and impossible solutions to ethnic civil wars. International
Security 20: 133–75.
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2007. Governance matters VI:
Governance indicators for 1996–2006. July. World Bank Policy Research Working Paper
No. 428.
Kymlicka, Will. 1995. Multicultural citizenship: A liberal theory of minority-rights. Oxford:
Clarendon.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny. 1999.
The quality of government. Journal of Law, Economics and Organization 15:
222–279.
Laponce, J. A. 2003. The case for ethno-linguistic federalism in multilingual societies. In The
Territorial Management of Ethnic Conflicts, ed. J. Coakley. London: Cass.
Lindstadt, Catharina, and Daniel Naurin. 2005. Transparency and corruption:
The conditional significance of a free press. Working Paper, Quality of Government
Institute.
Lijphart, Arend. 1977. Democracy in plural societies: A comparative exploration. New Haven:
Yale University Press.
———. 1988. Non-majoritarian democracy: A comparison of federal and consociation
theories. Publius: The Journal of Federalism 15: 3–15.
———. 1999. Patterns of democracy: Government forms and performance in thirty-six
countries. New Haven, CT: Yale University Press.
Mauro, Paolo. 1998. Corruption: Causes, consequences, and agenda for future research.
Finance and Development (International Monetary Fund), March: 11–14.
Mazzone, Jason. 2001. The social capital argument for federalism. Southern California
Interdisciplinary Law Journal 11: 27–62.
McGarry, John, and Brendan O’Leary, eds. 1993. The politics of ethnic conflict regulation.
London: Routledge.
McGarry, John, and Brendan O’Leary. 2007. Iraq’s Constitution of 2005: Liberal
consociation as political prescription. International Journal of Constitutional Law 5:
670–698.
Government Quality and Vertical Power-Sharing in Fractionalized States
21
Norris, Pipps. 2008. Driving democracy: Do power institutions work? New York: Cambridge
University Press.
O’Leary, Brendan, and John McGarry. 1995. Regulating nations and ethnic communities. In
Nationalism and Rationality, ed. Anton Breton, Gianluigi Galeotti, Pierre Salmon, and
Ronald Wintrobe, 245–90. New York: Cambridge University Press.
Ophuls, William P., and A. Stephen Boyan. 1992. Ecology and the politics of scarcity revisited:
The unraveling of the American dream. New York: W. H. Freeman.
Olson, M. 1965. The logic of collective action. Cambridge: Harvard University Press.
Ostrom, Elinor. 1990. Governing the commons. Cambridge: Cambridge University Press.
Persson, Torsten, and Guido Tabellini. 2003. The economic effects of constitutions. Cambridge,
MA: The MIT Press.
Persson, Torsten, Guido Tabellini, and Francesco Trebbi. 2003. Electoral rules and
corruption. Journal of European Economic Association 1: 958–989.
Posen, Barry. 1993. The security dilemma and ethnic conflict. Survival 35: 27–47.
Ray James, Lee. 2005. Constructing multivariate analyses. Conflict Management and Peace
Science 22: 277–292.
Rodrik, D. 2000. Institutions for high-quality growth: What they are and how to acquire
them. Studies in Comparative International Development 35: 3–31.
Rothstein, Bo, and Dietlind Stolle. 2007. The quality of government and social
capital: A thoery of political institutions and generalized trust. Working Paper 2007:2,
Quality of Government Institute.
Teorell, Jan, and Axel Hadenius. 2006. Does type of authoritarianism affect the prospects of
democracy? Exogenous shocks and contingent democratization. Working Paper 2006:3,
Quality of Government Institute.
Teorell, Jan, Sören Holmberg, and Bo Rothstein. 2008. The Quality of Government dataset,
version 15Nov06. Göteborg University: The Quality of Government Institute, http://
www.qog.pol.gu.se.
Treisman, Daniel. 2000. The causes of corruption: A cross-national study. Journal of Public
Economics 76: 399–457.
Verney, Douglas. 2004. Is ethnofederalism the solution or the problem? In Nationalism
after communism: Lessons learned, ed. Alina Mungiu-Pippidi and Ivan Krastev, 179–198,
New York: Central European University Press.
Watts, Ronald L. 1999a. The spending power in federal systems: A comparative study.
Kingston, Ontario: Institute of Intergovernmental Relations, Queen’s University.
———. 1999b. Comparing federal systems. 2nd ed. Kingston, Ontario: Queen’s University
and McGill-Queen’s University Press.
Weingast, Barry. 1997. The political foundations of democracy and the rule of law. American
Political Science Review 91: 245–263.